Download Eyewitnesses and the use and application of cognitive theory

CHAPTER 14 – EYEWITNESSES AND THE USE AND APPLICATION OF
COGNITIVE THEORY
Route Map
Introduction
14.1 INFORMATION RECALL
14.1.1 The Accuracy of Recall
Estimates of Physical Quantities
Stressful events
14.1.2 The Cognitive Interview
14.1.3 UK Interviewing Techniques
14.1.4 Descriptions of Faces
Patterns of Face Recall
Face Encoding
14.2 PERSON RECOGNITION
14.2.1 Unfamiliar Face Recognition
14.2.2 Identification Procedures
14.3 FACE CONSTRUCTION
14.3.1 Composite Systems: Past & Present
14.3.2 The Effectiveness of Facial Composite Systems
14.3.3 Internal & External Composite Features
14.3.4 Improving the Effectiveness of Composites
Combining memories
Facial Caricature
Improving Holistic Face Processing
Holistic Systems
14.4 SUMMARY: EVIDENCE & EYEWITNESSES
ROUTE MAP
Witnesses and victims may contain the only record of a crime and so their
evidence can be invaluable to a criminal investigation. This chapter reviews the
different types of evidence provided by these eyewitnesses. The initial focus is on
their ability to describe details of a crime and those involved, plus interviewing
techniques that attempt to maximise this information recall. Next, the procedures
used for identifying offenders are examined. In the final part, face construction
systems are discussed along with modern methods for improving their effectiveness.
1
INTRODUCTION
Eyewitnesses carry out a range of tasks to help the police bring a criminal to
justice. They first describe what happened during the crime and the persons involved.
Later, there may be reasonable grounds for the police to believe who is responsible
for the offence and this person then becomes a suspect. The police may put the
suspect into a lineup and ask eyewitnesses to see if they recognise the person they
saw. If the suspect is picked out, this is taken as evidence that the police have
arrested the correct person. They may then spend time trying to build a case to
convict the suspect, or dismiss it owing to a lack of evidence. Sometimes,
eyewitnesses provide testimony in a court of law.
The police may also use the given description to try to locate the offender
using computer searches of previously convicted persons. If a suspect cannot
otherwise be identified, eyewitnesses may construct a picture of the offender’s face, a
facial composite. The police will circulate this image in the newspapers, or on TV
crime programmes such as BBC Crimewatch, in the hope that a member of the public
will recognise it and phone in with a name. Sometimes, the police circulate a wanted
poster; see Figure 14.0 for an example. [FIGURE 14.0]
The focus of this chapter is on the various types of evidence from
eyewitnesses. There have been many cases of wrongful conviction, as illustrated later
by the conviction of Laszlo Virag, and these must be minimised. The chapter will be
guided by the wealth of psychological research that has attempted to understand and
improve the performance of human memory. It will be seen that eyewitness evidence
can be valuable for the detection and conviction of offenders.
2
14.1
INFORMATION RECALL
Eyewitnesses record a massive amount of information while observing a
crime. The focus of this section is on the recall of this information. The following
sections explore eyewitnesses’ recognition and face construction abilities.
14.1.1 The Accuracy of Recall
Yuille & Cutshall (1986) were first to formally evaluate eyewitness’s recall of
a real-life event. The crime involved a thief who stole guns and money from a gun
shop in Burnaby, Canada. In the street outside of the shop, shots were fired between
the shop owner and the thief, resulting in the death of the thief. Twenty-one
eyewitnesses were interviewed within two days of the event. Each person provided a
free account of what they saw and was questioned to expand upon their recall.
Thirteen of these observers also took part in a second, research interview between
four to five months later.
The descriptions given were surprisingly detailed and contained information
about actions, people and objects. Overall accuracy was about 80% correct in each of
these areas for both types of interview. The errors produced were fairly infrequent
and concerned: (a) actions of the thief, shop owner and other people present, (b)
descriptions of people’s height, weight, age, hair, colour of clothing and (c)
descriptions of objects.
Cutshall & Yuille (1989) investigated further shootings and armed robberies.
Although the quantity of eyewitness recall varied among crimes and among observers,
accuracy levels remained high for police and (later) research interviews, suggesting
3
that information in such situations is retained rather than forgotten over long periods
of time. Their result runs counter to the normal research findings which suggest that
information recall reduces progressively over time (Baddeley, 2004), but is perhaps
retained better in memory here owing to the serious nature of the crimes and also as
observers may rehearse information before being interviewed (Read, Hammersley,
Cross-Calvert & McFadzen, 1989). Again, inaccuracies involved descriptions of
people, especially colour of clothing and hair, and estimates of height, weight and
age. Note that these descriptions are often used forensically – they appear on wanted
posters, for example – but not all are errors of memory: height, weight and age are
arguably more to do with estimating physical quantities.
Estimates of physical quantities: Flin & Shepherd (1986) looked at our ability to
make judgements of height and weight. Over 500 people interacted with one of 14
target individuals, to obtain directions, and provided a height and weight estimate.
Some interesting patterns emerged. Firstly, people were fairly accurate in general:
estimates were inaccurate on average by about 8cm for height and 8kgs for weight.
Secondly, there was a tendency to give lower estimates for heavy and tall targets, but
higher estimates for light and short targets. The authors referred to this effect as a
regression towards the mean: extreme estimates are avoided in preference to those
that are more similar to the average or ‘norm’. Thirdly, taller participants tended not
to underestimate taller targets. Therefore, some participants made judgements based
on their own physical attributes.
Estimating another person’s age is quite a complex task owing to the range of
cues available (Ekman, 1978; see George & Hole, 2000, for a review), some of which
may not be helpful. For example, in an intriguing study reported by Sporer (1996),
4
Deusinger and Haase found that age estimates changed according to the clothing that
a target individual wore and the context in which he was seen. When appearing in an
aggressive scene wearing a red ‘windbreaker’ jacket, he was perceived to be four
years younger than the same person seen playing the role of an associate and wearing
a fine leather jacket. Research carried out using photographs of faces suggest we can
guess a person’s age correctly on average within about 5 years (e.g. Dehon & Bredart,
2001; George & Hole, 2000; Sörqvist & Eriksson, 2007). Our estimates of age, and
also of height and weight, are therefore fairly accurate, sufficient to provide a
reasonable indication for use in police investigations.
The Deusinger and Haase study highlights that a visual scene provides a
context that can influence our decisions processes. Our responses can even be
influenced by the wording of a question. In a classic study, Loftus & Palmer (1974)
showed participants a film of a road traffic accident. They were asked to describe
what happened and answer a series of questions about it. The particular question of
interest concerned the speed the vehicles were travelling when they contacted each
other. Some participants were asked alternative questions using hit, bumped, collided
or smashed instead of contacted. Average estimates of speed are shown in Figure
14.1.1 and increased from 31.8 mph for contacted to 40.5 mph for smashed. Simply
changing a verb can therefore dramatically affect what people report. [FIGURE
14.1.1]
In a second experiment, Loftus & Palmer’s participants saw a similar film and
one week later were asked whether they had noticed any broken glass. There was not,
in fact, any broken glass, but more people reported seeing it if they (1) gave higher
estimates of speed and (2) were asked whether the cars smashed into each other. The
study emphasises the need for caution when asking questions about an event, and,
5
without which, distortions may result (e.g. different estimates of speed). It is also
possible to create a false memory (broken glass). See Loftus (1992, 1997) and Sondhi
& Gupta (2007) for further examples.
Stressful events: Studies have examined the potentially negative influence of
anxiety. This is not an easy area of study, because of the large number of variables
involved, but the general finding is that people tend to recall less information in total,
and with lower accuracy, when they have observed an event under physiological
stress (Deffenbacher, Bornstein, Penrod & McGorty, 2004), although there are
exceptions (Hulse & Memon, 2006; Yuille & Cutshall, 1986). In a recent eyewitness
study, Valentine & Mesout (2009) examined the recall and recognition abilities of
visitors in the Horror Labyrinth at the London Dungeon. This is a situation where
people give their consent to be scared! Visitors who reported experiencing high
levels of physiological stress (state anxiety) were found to recall fewer correct details
and more incorrect details of a target individual; they also made fewer correct
identifications in a lineup.
A stressful event can affect memory retrospectively. Loftus & Burns (1982)
presented participants with either a violent or a non-violent film of a bank robbery and
asked them a series of questions about it. In the violent version, there was an
unexpected and mentally shocking conclusion involving a young person being shot in
the head. Fewer participants in the violent version (4.3%) correctly answered the
main ‘target’ question than those in the non-violent version (27.9%): what was the
number printed on a football jersey seen just before the end? Also, worse retention
was found for information seen in the final two minutes, an effect referred to as
retrograde amnesia, information memory loss. In addition, these effects applied to
6
upsetting but not unexpected (non-upsetting) events, further highlighting the role
played by personally-experienced stress.
The presence of a weapon can also impact upon recall. Participants in Loftus,
Loftus & Messo (1987) saw a series of slides involving a gun or a cheque (payment
for goods). Measurements of participants’ eye-movements revealed more fixations on
the gun rather than on the cheque, and each of these was for longer periods.
Participants in the condition involving the gun also answered fewer questions
correctly about the perpetrator. In such situations, there is a weapon focus: attention
tends to be drawn to the weapon and away from other objects, generally reducing the
information that is available for recall. Further, Loftus et al. found a weak effect on
recognition when a weapon was present, a reduction in correctly selecting the target
from a 12-person photo lineup. This is a finding reported in other laboratory research
(Steblay, 1992), but this may not be a reliable finding in the real world (Valentine,
Pickering & Darling, 2003).
14.1.2 The Cognitive Interview
Until the 1980s, the police followed a question-and-answer style interview to elicit
information from eyewitness: a standard interview. Sometimes, hypnosis-type
techniques were used to improve recall. Based on evidence that the standard
interview did not produce the most complete and accurate testimony, and that the
suggestive elements of hypnosis may result in false memories being implanted, Ron
Fisher, Edward Geiselman and their colleagues in the US developed an interviewing
‘system’ based on sound psychological principles (Geiselman, Fisher, MacKinnon &
Holland, 1985). This is known as the cognitive interview (CI).
7
The CI is based on the idea that the memory trace of a crime comprises of
different parts, or ‘components’. Some of these components relate to central,
important details, such as who did what to whom, and when, while others relate to
peripheral details such as what people were wearing, other objects present, the
lighting conditions, time of day, plus smells, sounds and personal feelings.
Theoretically, best recall occurs when as many components are activated as possible.
To achieve this objective, eyewitnesses are asked to think about the event and attempt
to recreate it in their mind. This is known as reinstating the context. Eyewitnesses
are instructed to report everything, even if seemingly irrelevant, as such details may
be important. They then provide a free recall of the event. However, we tend to
describe what we expect to have happened rather than what actually occurred. This
can be partly counteracted by recalling again from a different perspective; for
example, “What did the cashier see happen?” More generally, repeated recall
attempts take a different path through our memories and trigger previously unrecalled
components. This technique can also be carried out in a different temporal order, by
working backwards from the most recent event.
Geiseman et al. (1985) measured the effectiveness of the CI. Participants were
shown a film and then interviewed using a standard interview, a cognitive interview
or a hypnosis interview. As Figure 14.1.2 illustrates, the CI outperformed the
standard interview and was as good as hypnosis. [FIGURE 14.1.2]
In an attempt to further improve recall, Fisher et al. (1987) developed the
enhanced cognitive interview (ECI). The main stages are summarised in Table
14.1.2. As a high anxiety state can hinder recall, especially for victims of traumatic
events, police interviewers start by building rapport. Also, they clearly explain the
interview process and avoid interrupting during critical stages: free recall and context
8
reinstatement. Further, interviewers review the eyewitness’s account, which itself
provides another opportunity for recall. Laboratory tests (e.g. Fisher et al., 1987)
indicate that the ECI is even more effective than the CI (36.9% to 57.7% correct
details) with only a fairly small increase in incorrect recall (28%). Field trials also
support an increase in performance for the new interviewing techniques (e.g. Fisher,
Geiselman & Amador, 1989). [TABLE 14.1.2]
14.1.3 UK Interviewing Techniques
In the UK, the police use a PEACE interview, so called as it describes each
stage: Planning and preparation, Engage and explain, Account, Closure and
Evaluation. It includes the components of the ECI, as described above, as well as
further techniques for facilitating recall. These are called Conversation Management
(CM) and are based on the principle that an interview is, in essence, a complex
conversation between an interviewer and an eyewitness. In a conversation, peoples’
behaviour affects each other and, in an interview context, will determine to some
extent what a witness recalls. Important parts of CM involve listening skills, to
indicate that the interviewer is interested; at the simplest level, this includes the use of
‘minimal responses’ to show engagement, nodding and “aha” replies. It also involves
techniques to assist recall. For instance, if witnesses cannot remember a person’s
name, they can be asked to go through the alphabet thinking about the first letter of
the name, to help trigger the memory. Also important is transferring of control at
different stages; for example, before and after free recall. For a comprehensive
description of CM, see Milne & Bull (1999).
9
Police officers learn PEACE interviewing for use with witnesses over five
stages or tiers. Tier 1 is the basic level, which all police recruits receive, and this
normally takes 5 days. The training includes techniques of the enhanced CI; officers
gain much experience as part of normal operational duties. They may also attend
further training courses to provide additional skills, such as those necessary for
interviewing vulnerable and intimidated witnesses in Tiers 2 and 3. See Griffiths &
Milne (2005) for a review of interviewing tiers.
14.1.4 Descriptions of Faces
Eyewitnesses also provide a description of the offender’s face. As with person
descriptions in general, this information is used for public appeals and computer
searches. Facial descriptions play an additional role for constructing lineups and facial
composites, as discussed later. Before reading this section, first complete Activity
Box 1. [ACTIVITY BOX 1]
Human faces present an interesting challenge for our perceptual system: they
all share the same basic design. All have individual facial features – two eyes, two
brows, a nose, a mouth, two ears – and each of these is placed in roughly the same
configuration, or position on the face: brows above eyes, nose between eyes and
mouth, etc. Men, women and people from different races have these properties in
common. It is the differences in this basic template that gives rise to individuality.
Detecting offenders is likely to be more effective if eyewitnesses’ descriptions
contain aspects that are unusual or distinctive. Perhaps there was a small scar, tattoo
or blemish; a facial feature may be unusual in some way: small eyes, large eyebrows
or unusual shaped ears. The facial configuration can be distinctive and provide
10
important perceptual cues; some people, for example, have widely-spaced eyes.
Facial distinctiveness, due to unusual features or an unusual configuration, is an
important aspect of face perception as it can enhance recognition accuracy (Shapiro &
Penrod, 1986).
Patterns of face recall: Laughery, Duval & Wogalter (1986) asked participants to
describe an unfamiliar face from their memory. They found that more descriptions
were given for features in the upper than the lower half of the face, especially for the
hair and eyes. Most adjectives described the size and shape of features, but were not
specific to faces; examples include large eyes, small nose, thin lips and dark
eyebrows. Face description is not an easy task, although the hair was reported to be
the easiest feature to describe.
Ellis, Shepherd & Davies’s (1980) participants were shown one of two
unfamiliar target faces and recalled details immediately, or after one of three time
delays: one hour, one day or one week. As with Laughery et al., most information
was recalled for the hair and eyes, and then for the general structure of the face
including the shape, length and size of head. Ellis et al. found that the quantity of
information recalled reduced as the delay increased: it was greatest when recall
occurred immediately and lowest after a week. This is a typical effect that occurs for
verbal information in general (e.g. Baddeley, 2004).
A second group of participants matched the descriptions to a photograph of the
targets. Delay had a progressively detrimental effect: descriptions were matched less
accurately when produced after longer time intervals. Therefore, the quality of
information remembered about an unfamiliar face reduces over time. Eyewitnesses
often recall an offender’s face several days, or even longer, after a crime and so are
11
likely to have forgotten information that would have been useful for identification.
Even using cognitive interview techniques, unfortunately, the majority of witnesses
only provide a sketchy description of the face.
Face encoding: The method we use to remember or encode a face affects our ability
to describe it. Some eyewitnesses correctly suppose that they may be asked questions
about an offender’s face and so look in detail at the individual features, making
‘mental notes’ to themselves about sizes, shapes and colours. This is known as a
feature encoding. Research has found that the quality of descriptions are better
following such an encoding than when the face has been attended to more naturally,
as part of communication, or by making personality judgements about it (Wells and
Turtle, 1988). Observers who use the latter holistic encoding are better at whole-face
type tasks such as identifying a face from alternatives (Wells & Hryciw, 1984). The
effect of encoding provides insight into how our cognitive system carries out different
tasks: face description involves individual features, while face recognition involves a
more overall analysis of the face. It also suggests that eyewitnesses will produce a
more accurate and a more complete description under feature encoding; note that the
majority of us do not do this due to a natural tendency to encode faces holistically
(Olsson & Juslin, 1999).
14.2
PERSON RECOGNITION
There are many cues we use to recognise a person: body shape and build,
voice and accent, gait, face, behaviour and clothing. But, which of these are
12
important for recognising a person? Burton, Wilson, Cowan & Bruce (1999) used
video footage from a surveillance camera of members of Psychology staff at Glasgow
University. Participants were shown original footage, or where editing had obscured
the body, gait or face; examples are shown in Figure 14.2. People familiar with the
staff made correct identifications at over 80% on average for each presentation format
except the one with the face concealed (less than 40%). The study shows that the face
is the most valuable cue for recognising a familiar person, someone we have seen
many times. [FIGURE 14.2]
We recognise familiar faces with almost perfect accuracy, even from different
viewpoints, under different lighting conditions and in various environments (contexts)
(Bruce & Young, 1998). Recognition is very accurate, even under less than optimal
conditions such as a grainy wedding photograph or low quality CCTV (e.g. Lander,
Bruce & Hill, 2001). For this reason, TV crime programmes show CCTV images: so
long as the camera has captured an un-obscured image of the offender’s face,
someone who knows the person should be able to recognise him or her from it.
14.2.1 Unfamiliar Face Recognition
Some early research studies seemed to suggest that we are also good at
recognising unfamiliar faces (e.g. Goldstein & Chance, 1971). These studies used a
recognition memory task (RMT): participants were presented with a series of faces
during a study phase and were asked to recognise these same items when mixed in
with other faces during a test phase. All faces were previously unfamiliar to
participants. The problem is that we can do this exercise using properties of the
picture: how a person looked into the camera, the specific lighting, their expression,
13
the background scene, etc. Showing a different photograph between study and test
explores our ability to process facial structure (identity). Using a RMT, Bruce (1982)
found accuracy rates at about 90% correct with identical photos, but this was lower
when test phase faces were changed by head pose or expression, and lowest when
changes occurred to pose and expression. Her results are provided in Figure 14.2.1.
Further research has confirmed that performance deficits occur following changes to
the background scene, hair and lighting conditions (Cutler, Penrod & Martens, 1987;
Davies & Milne, 1982; Hill & Bruce, 1996). Our performance also becomes
measurably worse when an observer is not used to discriminating faces from a
particular race. This is known as a cross-race effect (see Meissner & Brigham, 2001,
for a review). [FIGURE 14.2.1]
Ellis, Shepherd & Davies (1979) looked at how we process different facial
regions. Participants were shown well-known (celebrity) photographs of internal
features – the central part of the face containing eyes, brows, nose and mouth – or
external features – the region including hair, face shape and ears. Other participants
saw complete faces. Naming of intact faces was about 80% correct, internal features
about 50% and external features about 30%. Next, a RMT was used with unfamiliar
faces, and internal and external features were recognised equally well. The work
demonstrates a greater reliance on the internal features for recognising a familiar face,
but on the external features when familiarity is low.
In police investigations, an offender’s face is normally seen just once and so
the external parts will have greater importance to an eyewitness than if the person was
well-known. Further research has shown that hair is one of the most important
external features for unfamiliar face perception (Bruce et al., 1999; Ellis, 1986; Frowd
14
& Hepton, 2009). This means that offenders can reduce the chances of being
recognised simply by changing the colour, style and/or length of their hair.
Recognition accuracy is also strongly affected by facial similarity.
Participants in Davies, Shepherd & Ellis (1977) first sorted 100 faces into piles of
similar appearance. The researchers then constructed ‘lineups’ containing faces that
had been sorted together frequently or infrequently, and presented these to further
participants in an identification task. Overall accuracy was about 75% correct, but the
vast majority of errors occurred for the arrays constructed from frequently sorted
items. The work demonstrates that faces judged similar to each other are also likely
to be confused. Therefore, eyewitnesses are more likely to make errors when facial
similarity is high. Unfortunately, confusions also occur when similarity is much
lower: see Focus Point 14.2.1. [FOCUS POINT 14.2.1]
14.2.2 Identification Procedures
Eyewitnesses in the UK participate in two types of identification procedure
(PACE Code D, 2008). Firstly, the police may escort eyewitnesses to public places to
see if they can spot the offender. This is known as a group identification. The
second procedure involves a lineup and is used when the police have a suspect. One
way to do this is to place the suspect in a line with at least 8 other individuals (‘foils’)
and ask the eyewitness to pick out the offender if present. This is known as an
identification parade. Alternatively, the suspect and foils may be filmed and shown
to eyewitnesses as a sequence of video clips, a video identification. There is a system
in the UK to do this called VIPER, or Video Identification Parade Electronic
Recording (Brace, Pike, Kemp & Turner, 2009).
15
VIPER is an example of a sequential lineup. It is, however, not the most
common format worldwide. Normally, eyewitnesses see all lineup members at the
same time, a simultaneous lineup. There has been considerable debate about which
presentation method is the best (e.g. Meissner, Tredoux, Parker & MacLin, 2005), but
recent evidence suggests that the simultaneous format is slightly better (McQuistonSurrett, Malpass & Tredoux, 2006); note that there are also some very promising,
novel lineup procedures (e.g. Levi, 1998, 2006).
Considerable effort has been spent generally trying to understand the factors
involved with lineups, with the aim of maximising the chances of an eyewitness
identifying a guilty suspect (the person who actually committed the offence) but
minimising the identification of innocent suspects (there are other reasons, however,
some based on legal systems different to ours; see Wells, 1985). One such factor, for
example, is the quality and quantity of a person’s description of a face and his or her
accuracy in a lineup, a description-identification relationship. The relationship turns
out to be fairly complicated, but weak effects do appear to exist; the reader interested
in this is directed to Meissner, Sporer & Susa (2008) for a review.
It is crucial that lineups are fair. If suspects stand out in some way, the
chances of them being picked out will increase. When an innocent suspect is selected,
police time is wasted and the chances of wrongful conviction will increase. One way
to improve the effectiveness of lineups is to increase their size. All else being
constant, this reduces an innocent suspect being selected by chance alone; for
example, the chance is 1 in 10 (a probability of 0.1) for a 10-person lineup but half
that for a 20-person lineup (0.05). Research suggests that increasing lineup size does
not appear to reduce identification rates of guilty suspects, fortunately, even for very
large lineups (Levi, 2002, 2007).
16
A well-intentioned, but nonetheless flawed procedure is to select members of a
lineup that look similar to a photograph of the suspect. The problem is that, if shown
a series of faces, we have a tendency to form an average of the set in our mind. The
average is known as a prototype and becomes a memory with which we will respond
as if the face had really been seen (Cabeza, Bruce, Kato & Oda, 1999). So, if a lineup
is constructed from a suspect’s photograph, the suspect becomes the prototype and
witnesses will tend to select it! A better procedure is to select members based on a
verbal description of the suspect given by an eyewitness (e.g. Brigham, Ready &
Spier, 1990): this is one reason to ask them for a description in the first place.
One way to test the fairness of a lineup is to administer a mock witness
procedure (Doob & Kirshenbaum, 1973), which can detect prototype effects and
distinctive lineup members. To do this, a group of volunteers (people who are not
familiar with the suspect) are given the eyewitness’s description, shown the lineup
and asked to select who they think is the suspect. Any lineup member who is picked
out more often than chance alone provides evidence that the lineup is biased – and,
should be re-constructed. Using this procedure, video identification has been found to
be less biased (Valentine & Heaton, 1999) and more ergonomic (Brace et al., 2009;
Kemp, Pike & Brace, 2001) than live lineups (identification parades).
14.3
FACE CONSTRUCTION
Witnesses often participate in lineup type identification procedures many
months into an investigation. In the absence of a suspect, they may help earlier on by
constructing a visual likeness of the offender’s face. These pictures are known as
17
facial composites. They are produced by a number of methods or ‘systems’ after a
cognitive interview, and are shown to police officers or members of the public for
identification. The use of composites therefore involves two types of perceptual
process: unfamiliar face perception to construct the face (by eyewitnesses) and the
familiar type to recognise it (police officers and the public). Given the potential
importance of these images for the detection of offenders, composites have been the
focus of considerable research and development.
14.3.1 Composite Systems: Past & Present
The earliest method of face construction was a sketch artist. This person
would be skilled in portraiture and would draw the face using pencils or crayons. The
first UK system for police operatives with less artistic skill was Photofit. The method
used individual reference features printed onto rigid card; eyewitnesses would select
the best matching hair, eyes, nose, etc., with each part being slotted together as if
assembling a jigsaw. Photofit has been extensively evaluated and has led to the
software systems that the police currently use; for a review, see Ellis & Shepherd
(1996).
One problem with Photofit was that eyewitnesses were required to select from
sets of isolated facial features. We never naturally do this: eyes, for example, are
always seen in a face with a nose, brows, a mouth, etc. It turns out that we are more
accurate at selecting features when they are embedded in an intact face (e.g. Davies &
Christie, 1982; Tanaka & Farah, 1993). This idea has been applied to modern
software systems as part of a ‘cognitive’ approach to face building: witnesses are
18
presented with features that are switched in and out of a single, intact face; the method
produces more identifiable composites (Skelton & Frowd, manuscript in preparation).
A second problem was a limitation in the range of features: there were simply
not enough examples to build composites of all offenders. This issue was addressed
by photographing a large number of people, and then classifying and storing their
facial parts in a database; also, computer graphics technology permitted each feature
to be resized and positioned freely, vastly expanding system expressivity.
Unfortunately, there was now too many basic examples to show to witnesses! In the
UK PRO-fit software system, for example, 219 noses are stored in the White Male
database, but a manageable set of 14 are ‘long’ and ‘straight’. As with lineups, a
witness’s description is now an important prerequisite, in this case to specify subsets
of features within the database. A cognitive interview is again administered for
eliciting face recall, with the main techniques being rapport building, context
reinstatement, free recall and cued recall; a ‘verbal description’ sheet is used by police
operatives for writing down the details (see Activity Box).
14.3.2 The Effectiveness of Facial Composite Systems
Frowd et al (2005b) evaluated Photofit, along with its descendents, E-FIT and
PRO-fit, sketch and an early system in development, EvoFIT, which is described in
section 14.3.4. Participants looked at a good quality target photograph, a face they
did not recognise, waited three-to-four hours and then followed procedures used with
‘real’ witnesses. This involved a CI, to elicit a description of the face, and use of the
composite system as specified by the manufacturers to produce the best likeness
possible. The targets were celebrity faces, half of which were distinctive, and half
19
more average in appearance. Example images from the study can be seen in Figure
14.3.2.A [FIGURE 14.3.2.A]
The composites were shown to other people to name (Figure 14.3.2.B). The
two UK software systems, E-FIT and PRO-fit, performed equivalently at 18% correct
overall. Note that this level is only fairly good, but was significantly better than the
older Photofit, at 6%. Naming was low for sketch, at 9%, and very low for EvoFIT, at
1%. The study revealed that composites of distinctive targets were named about three
times better overall than composites of average targets. This target distinctiveness
advantage is also found when recognising photographs of faces (e.g. Shapiro &
Penrod, 1986); here, it suggests that an offender will be identified better from a
composite if his or her face has an unusual appearance. [FIGURE 14.3.2.B]
Frowd et al. (2005a) asked participants to wait 2 days between seeing a target
and constructing a face, the typical delay for real witnesses. Sketches were correctly
named about the same as before, at 8%, but E-FIT and PRO-fit were less than 1%
overall! This indicates that the ‘feature’ systems are failing to produce recognisable
images when deployed after a realistic 2 day delay: the exception is sketch, and even
then performance is not great. Using a similar design with a 2 day delay, other
projects have found similarly disappointing performance for software feature systems
(Frowd et al., 2007b, 2007d, in press).
14.3.3 Internal and External Composite Features
Frowd et al. (2007a) attempted to understand why naming should be very poor
for software composites constructed after a 2 day delay. Similar to Ellis et al’s (1979)
study using photographs of faces, Frowd et al. examined two regions of the face: the
20
internal and external features; see Figure 14.3.3 for examples. A naming task could
not be used as levels would be too low for analysis; instead a sorting procedure was
administered that required participants to match complete and part-face composites to
target photographs. Scores were 33% correct for the complete and external features
composites, but only 20% for composites of internal features. The study suggests that
low composite naming was a result of poor quality internal features; also, that if this
region of the face could be constructed more accurately, then better recognition would
emerge for the image as a whole. [FIGURE 14.3.3]
14.3.4 Improving the Effectiveness of Composites
The following techniques have successfully improved composite naming rates.
Each is in current police use.
Combining different memories: Bruce et al. (2002) considered the situation where
multiple witnesses had seen the same offender. If each observer were to construct a
composite, each image would be different, but all would share characteristics of the
offender. This can be clearly seen in Figure 14.3.4.A. Images would also contain
errors. The authors’ hypothesised that if the composites were combined, then some of
these errors would be cancelled out and the resulting ‘average’ or morphed composite
would be more accurate. Their hypothesis was confirmed in a series of experiments:
a morphed composite was judged to have a better likeness than an average individual
composite and, in an identification task, was selected as often as the best individual
image. [FIGURE 14.3.4.A]
21
Prior to Bruce et al’s work, the UK police guidelines did not permit more than
one composite to be constructed of the same offender in a multiple witness scenario.
Since then, the guidelines have been updated to allow construction of multiple
images, for the purpose of producing a morphed composite for public appeals
(NPIA/ACPO, 2009). This image would feature on a wanted poster (such as on
Figure 14.0) in preference to any individual composite.
When there is only one witness to a crime, however, morphing techniques
cannot be used. In this more usual situation, other methods are required for
composites to be effective. Some of these are described below.
Facial caricature: Different observers, even after seeing a target face for the same
length of time, construct different-looking images (see Figure 14.3.4.A). Individual
differences are also observed when people attempt to recognise composites: some of
us are very good at naming them, while others struggle. It is likely that composites
tend to be quite bland in appearance, and also contain ‘error’, which together make
their recognition generally difficult. Frowd et al. (2007c) looked at enhancing
distinctive information in composites and reducing error. Participants observed a
composite being caricatured, where the distinctive aspects were exaggerated, and
then anti-caricatured, where these same aspects were made to look more average.
This manipulation was seen over a 21 frame sequence, as illustrated in Figure
14.3.4.B. [FIGURE 14.3.4.B]
The sequence is very effective at triggering the memory of the relevant
identity. The effect extends to all construction methods – ‘feature’, sketch and
‘holistic’ – but is greatest for the worst quality composites: those produced after a
long delay. For these images, average naming was found to increase tenfold as a
22
result of watching the sequence, from 3% to 30%. It is most conveniently seen using
an animated GIF format on a wanted person’s webpage or on TV.
Improving holistic face processing: A cognitive interview (CI) is used to recall
information about a face, to allow subsets of features to be located within a composite
system for presentation to eyewitnesses. However, eyewitnesses engage in face
recognition when they select facial features. This is the reason why the ‘cognitive’
approach is used in modern systems (as described in 14.3.1). But, this idea can be
taken a stage further, by enhancing recognition ability via personality attribution.
This was first demonstrated using photographs of faces. Berman & Cutler (1989)
demonstrated that people were better able to recognise unfamiliar faces if they had
made a series of personality judgements about them than if they had made facial
feature judgements.
Frowd et al. (2008b) asked one group of participants to construct a face as
normal, after a CI, while another group did the same but were asked to think about the
personality of the face (similar to the CI’s free recall component) and then make a
series of whole-face judgements (cued recall). Examples include: friendliness,
honesty and distinctiveness. Naming of the resulting composites were 9% after the
CI, but 41% after the ‘holistic’ CI (H-CI), a four fold increase. The additional
procedure of the H-CI takes a few minutes to administer but is very effective for
improving the quality of an individual composite.
Holistic systems: All of the techniques described above run into difficulty when a
witness’s description is very sketchy, as feature subsets cannot be located. This can
happen for briefly-occurring or unexpected crimes (normally where witnesses do not
23
try to remember the face). In spite of this, they report that they could recognise the
face if seen again; they could, for example, participate in a lineup if only a suspect
could be located. Such witnesses appear to have a memory of the face that, while not
in a describable form, could potentially be accessed via a recognition procedure. This
idea is supported by research to suggest that face recognition ability remains stable for
several weeks after an event (Shepherd, 1983), unlike face recall. Several researchers
have been developing so-called ‘holistic’ systems to help these eyewitnesses. In the
UK, there are two such commercial systems: E-FIT-V (Gibson, Solomon & PallaresBejarano, 2003) and EvoFIT (Frowd, Bruce & Hancock, 2008a).
In collaboration with Vicki Bruce and Peter Hancock, I have been developing
EvoFIT for about 10 years. Witnesses are presented with arrays of complete faces
from which to select a number of faces most resembling the perpetrator’s face. The
selected items are bred together, to combine characteristics, and produce more faces
for selection. Repeating the process a few times allows the faces to become more
similar to each other and more similar to the face in the memory of the observer. The
item with the best likeness is ultimately saved as the ‘composite’. Therefore,
witnesses engage their face recognition ability, by selecting items that look overall
similar to the intended target, to allow a composite to be ‘evolved’. EvoFIT is a
working example of Charles Darwin’s idea of ‘evolution by artificial selection’. In
practice, the process is refined: witnesses look at face arrays that change by shape –
the shape and position of features – and then by texture – the colour of the eyes,
brows and overall skin tone. [FIGURE 14.3.4.C]
In tests using the standard police procedure and a 2 day delay, this version of
EvoFIT produces more identifiable composites than those from a typical ‘feature’
system, but naming levels are still fairly low, at 11% (Frowd et al., 2007b, in press).
24
Two developments have improved performance. The first was to ‘blur’ the external
facial features at the start. This helps a witness to focus on the central part of the face
that is important for later recognition of the composite by another person (Frowd et
al., 2008c) – see Figure 14.3.4.C. The other was to allow enhancement of an
‘evolved’ face using a number of psychologically-useful scales (Frowd et al., 2006).
Thus, observers can make changes to the face by age, weight, masculinity,
friendliness, and other holistic properties of the face; see Figure 14.3.4.D for an
example. In a recent evaluation, these two improvements were very effective at
evolving a face after 2 days: correct naming was 25% correct compared to 5% for the
‘feature’ system (Frowd et al., in press). [FIGURE 14.3.4.D]
An important part of research is to verify performance outside of the
laboratory, as part of fieldwork. This was seen for the cognitive interview (e.g. Fisher
et al., 1989). EvoFIT has been audited in police field trials. In spite of difficulties in
controlling for variables in such research (e.g. target encoding, exposure time and
delay) and, as part of a six month trial period, Lancashire police have reported
EvoFIT to be valuable in 20% of cases, Derbyshire 30%. Therefore, results are
similar both in the laboratory and in the hands of the intended user.
14.4
SUMMARY: EVIDENCE AND EYEWITNESSES
Evidence collected from eyewitnesses is just one of many different sources
available in a criminal investigation. Criminals may unknowingly leave their DNA at
a crime scene, have touched objects, left fingerprints; even fibres from clothing may
fall and be collected for analysis. These sources may also produce false leads, since
25
we naturally leave behind such physical materials: hairs fall off, we make footprints
and CCTV cameras record our presence. As with the case of Laszlo Virag and others,
any one piece of evidence is not sufficiently reliable as a basis on which to convict.
The Crown Prosecution Service in the UK is acutely aware of this issue and will not
allow a case to be brought before the courts if it is based on insufficient or potentially
unreliable evidence.
In spite of the unreliable nature of human memory, eyewitnesses can be very
valuable to a police investigation. There are many situations where they are the only
observers, and so their memory of what happened is of paramount importance. Each
of our senses produce memories that are potentially valuable, even personal feelings!
Eyewitnesses can provide a range of evidence to assist in an investigation. They can
provide descriptions of events and people, take part in identification procedures and
construct a likeness of the offender’s face. Each area of their evidence has potential
problems, but research has attempted to limit bias and maximise the value of a
memory. Each piece of evidence is potentially unreliable but, when combined, can
provide a reliable system for identifying and convicting criminals.
(6,886 words.)
26
References
ACPO (2009). Facial Identification Guidance. Produced by the National Policing
Improvement Agency. Unpublished document.
A. Baddeley (2004). Human memory: a user’s guide. New York: Firefly Books.
Berman, G.L., & Cutler, B.L. (1998). The influence of processing instructions at
encoding and retrieval on face recognition accuracy. Psychology, Crime & Law, 4,
89-106.
Brace, N.A., Pike, G.E., Kemp, R.I., & Turner, J. (2009). Eye-witness identification
procedures and stress: a comparison of live and video identification parades.
International Journal of Police Science & Management, 11 183-192.
Brigham, J. C., Ready, D. J., & Spier, S. A. (1990). Standards for evaluating the
fairness of photograph lineups. Basic and Applied Social Psychology, 11, 149-163.
Bruce, V. (1982). Changing faces: Visual and non-visual coding processes in face
recognition. British Journal of Psychology, 73, 105-116.
Bruce, V., Henderson, Z., Greenwood, K., Hancock, P.J.B., Burton, A.M., & Miller,
P. (1999). Verification of face identities from images captured on video. Journal of
Experimental Psychology: Applied, 5, 339 360.
27
Bruce, V., Ness, H., Hancock, P.J.B, Newman, C., & Rarity, J. (2002). Four heads are
better than one. Combining face composites yields improvements in face likeness.
Journal of Applied Psychology, 87, 894-902.
Bruce, V. & Young, A. (1998). In the eye of the beholder: The science of face
perception. New York: Oxford University Press.
Burton, A.M., Wilson, S., Cowan, M., & Bruce, V. (1999). Face recognition in poor
quality video: evidence from security surveillance. Psychological Science, 10, 243248.
Cabeza, R., Bruce, V., Kato, T., & Oda, M. (1999). The prototype effect in face
recognition: extension and limits. Memory & Cognition, 27, 139-151.
Cutler, B. L., Penrod, S. D., & Martens, T. K. (1987). Improving the reliability of
eyewitness identifications: Putting context into context. Journal of Applied
Psychology, 72, 629-637.
Cutshall, J.L., & Yuille, J.C. (1989). Field studies of eyewitness memory of actual
crime scenes. In Raskin, D.C. (Ed.), Psychological methods in criminal investigation
and evidence (pp. 97-124). New York: Springer.
Davies, G.M. (1983). Forensic face recall: the role of visual and verbal information.
In S.M.A. Lloyd-Bostock and B.R. Clifford (Eds.). Evaluating witness evidence (pp.
103-123). Chichester: Wiley.
28
Davies, G.M., & Christie, D. (1982). Face recall: an examination of some factors
limiting composite production accuracy. Journal of Applied Psychology, 67, 103-109.
Davies, G.M., & Griffiths, L. (2008). Eyewitness Identification and the English
Courts: A Century of Trial and Error. Psychiatry, Psychology and Law. 15, 435–449.
Davies, G.M., & Milne, A. (1982). Recognizing faces in and out of context. Current
Psychological Research, 2, 235-246.
Davies, G.M., Shepherd, J.W., & Ellis, H.D. (1977). Similarity effects in face
recognition. American Journal of Psychology, 92, 507-523.
Deffenbacher, K.A., Bornstein, B.H., Penrod. S.D, & McGorty, E.K. (2004). A metaanalytic review of the effects of high stress on eyewitness memory. Law and Human
Behavior, 28, 687-706.
Dehon, H., & Bredart, S. (2001). An ‘other-race’ effect in age estimation from faces.
Perception, 30, 1107 – 1113.
Devlin, P. (1976). Report to the Secretary of State for the Home Department of the
Departmental Committee on Evidence of identification in criminal cases. London:
HMSO.
29
Doob, A.N., & Kirshenbaum, H.M. (1973). Bias in Police line-ups - partial
remembering. Journal of Police Science and Administration, 1, 287-293.
Ekman, P. (1978). Facial signs: facts, fantasies and possibilities. In T. Sebeok (Ed.),
Sight, sound and sense (pp. 124-156). Bloomington: Indiana University Press.
Ellis, H. D. (1986). Face recall: A psychological perspective. Human Learning, 5, 1-8.
Ellis, H.D., Shepherd, J.W., & Davies, G.M. (1980). The deterioration of verbal
descriptions of faces over different delay intervals. Journal of Police Science and
Administration, 8, 101-106
Ellis, H.D., & Shepherd, J.W. (1996). Face recall methods and problems. In S.L.
Sporer, R.S. Malpass & G. Koehnken (Eds.). Psychological issues in eyewitness
identification. Hillsdale, NJ: Lawrence Erlbaum. Ch. 5.
Ellis, H.D., Shepherd, J.W., & Davies, G.M. (1979). Identification of familiar and
unfamiliar faces from internal and external features: some implications for theories of
face recognition. Perception, 8, 431-439.
Fisher, R.P., Geiselman, R.E., & Amador. M. (1989). Field test of the cognitive
interview: Enhancing the recollection of actual victims and witnesses of crime.
Journal of Applied Psychology, 74, 722-727.
30
Fisher, R.P., Geiselman, R.E., Raymond D.S., Jurkevich, L.M., & Warhaftig, M.L.
(1987). Enhancing enhanced eyewitness memory: Refining the cognitive interview.
Journal of Police Science and Administration, 15, 291-297.
Flin, R.H., & Shepherd, J.W. (1986). Tall stories: Eyewitnesses' ability to estimate
height and weight characteristics. Human Learning: Journal of Practical Research &
Applications, 5, 29-38.
Frowd, C.D., Bruce, V., & Hancock, P.J.B. (2008a). Changing the face of criminal
identification. The Psychologist, 21, 670-672.
Frowd, C.D., Bruce, V., McIntyre, A., & Hancock, P.J.B. (2007a). The relative
importance of external and internal features of facial composites. British Journal of
Psychology, 98, 61-77.
Frowd, C.D., Bruce, V., McIntyre, A., Ross, D., Fields, S., Plenderleith, Y., &
Hancock, P.J.B. (2006). Implementing holistic dimensions for a facial composite
system. Journal of Multimedia, 1, 42-51.
Frowd, C.D., Bruce, V., Ness, H., Bowie, L., Thomson-Bogner, C., Paterson, J.,
McIntyre, A., & Hancock, P.J.B. (2007b). Parallel approaches to composite
production. Ergonomics, 50, 562-585.
31
Frowd, C.D., Bruce, V., Ross, D., McIntyre, A., & Hancock, P.J.B. (2007c). An
application of caricature: how to improve the recognition of facial composites. Visual
Cognition, 15, 1-31.
Frowd, C.D., Bruce, V., Smith, A., & Hancock, P.J.B. (2008b). Improving the quality
of facial composites using a holistic cognitive interview. Journal of Experimental
Psychology: Applied, 14, 276 – 287.
Frowd, C.D., Carson, D., Ness, H., McQuiston, D., Richardson, J., Baldwin, H., &
Hancock, P.J.B. (2005a). Contemporary Composite Techniques: the impact of a
forensically-relevant target delay. Legal & Criminological Psychology, 10, 63-81.
Frowd, C.D., Carson, D., Ness, H., Richardson, J., Morrison, L., McLanaghan, S., &
Hancock, P.J.B. (2005b). A forensically valid comparison of facial composite
systems. Psychology, Crime & Law, 11, 33-52.
Frowd, C.D., & Hepton, G. (2009). The benefit of hair for the construction of facial
composite images. British Journal of Forensic Practice, 11, 15-25.
Frowd, C.D., McQuiston-Surrett, D., Anandaciva, S., Ireland, C.E., & Hancock,
P.J.B. (2007d). An evaluation of US systems for facial composite production.
Ergonomics, 50, 1987–1998.
Frowd, C.D., Park., J., McIntyre, A., Bruce, V., Pitchford, M., Fields, S., Kenirons,
M. & Hancock, P.J.B. (2008c). Effecting an improvement to the fitness function. How
32
to evolve a more identifiable face. In A. Stoica, T. Arslan, D. Howard, T. Higuchi,
and A. El-Rayis (Eds.) 2008 ECSIS Symposium on Bio-inspired, Learning, and
Intelligent Systems for Security (pp. 3-10). NJ: CPS. (Edinburgh).
Frowd, C.D., Pitchford, M., Bruce, V., Jackson, S., Hepton, G., Greenall, M.,
McIntyre, A., & Hancock, P.J.B. (in press). The psychology of face construction:
giving evolution a helping hand. Applied Cognitive Psychology.
Geiselman, R.E., Fisher, R.P., MacKinnon, D.P., & Holland, H.L. (1985). Eyewitness
memory enhancement in the police interview: Cognitive retrieval mnemonics versus
hypnosis. Journal of Applied Psychology, 70, 401-412.
George, P.A., & Hole, G.J. (2000). The role of spatial and surface cues in the ageprocessing of unfamiliar faces. Visual Cognition, 7, 485-509.
Gibson., S.J., Solomon, C.J., & Pallares-Bejarano, A. (2003). Synthesis of
photographic quality facial composites using evolutionary algorithms. In R. Harvey
and J.A. Bangham (Eds.) Proceedings of the British Machine Vision Conference (pp.
221-230).
Goldstein, A.G., & Chance, J.E. (1971). Visual recognition memory for complex
configurations. Perception & Psychophysics, 9, 237-241.
33
Griffiths, A., & Milne, R. (2005). Will it all end in tiers? Police interviews with
suspect in Britain. In T.Williamson (Ed.), Investigative interviewing. Rights, research,
regulation. Devon:Willan Publishing.
Hill, H., & Bruce, V. (1996). Effects of lighting on the perception of facial surfaces.
Journal of Experimental Psychology: Human Perception and Performance, 22, 9861004.
Hulse, L.M. & Memon, A. (2006). Fatal impact? The effects of emotional arousal and
weapon presence on police officers’ memories for a simulated crime. Legal and
Criminological Psychology, 11, 313–325.
Kemp, R.I., Pike, G.E., & Brace, N.A. (2001). Video-based identification procedures:
Combining best practice and practical requirements when designing identification
systems. Psychology, Public Policy, and Law, 7, 802-807.
Lander, K., Bruce, V., & Hill, H. (2001). Evaluating the effectiveness of pixelation
and blurring on masking the identity of familiar faces. Applied Cognitive Psychology,
15, 101-116.
Laughery, K.R., Duval, C., & Wogalter, M.S. (1986). Dynamics of facial recall, in
Ellis, H.D., Jeeves, M.A., Newcombe, F., & Young, A. (Eds.). Aspects of face
processing (pp. 373-387). Dordrecht: Martinus Nijhoff.
34
Levi, A.M. (1998). Protecting innocent defendants, nailing the guilty: a modified
sequential lineup. Applied Cognitive Psychology, 12, 265-275.
Levi, A.M. (2002). Up to forty: Lineup size, the modified sequential lineup, and the
sequential lineup. Cognitive Technology, 7, 39-46.
Levi, A.M. (2006). An analysis of multiple choices in MSL lineups, and a
comparison with simultaneous and sequential ones. Psychology, Crime & Law, 12,
273-285.
Levi, A.M. (2007). Research note: evidence for moving to an 84-person photo
lineup. Journal of Experimental Criminology, 3, 377-391.
Loftus, E.F. (1992). When a lie becomes memory’s truth. Current Directions in
Psychological Science, 1, 121-123.
Loftus, E. F. (1997). Creating false memories. Scientific American, 277, 70-75.
Loftus, E.F., & Burns, T.E. (1982). Mental shock can produce retrograde amnesia.
Memory & Cognition, 10, 318-323.
Loftus, E.F., Loftus, G.R., & Messo, J. (1987). Some facts about weapon focus. Law
and Human Behavior, 11, 55-62.
35
Loftus, E. F., & Palmer, J. C. (1974). Reconstruction of automobile destruction: An
example of the interaction between language and memory. Journal of Verbal
Learning and Behavior, 13, 585-589.
McQuiston-Surrett, D., Malpass, R. S., & Tredoux, C. G. (2006). Sequential vs.
simultaneous line-ups: A review of methods, data, and theory. Psychology, Public
Policy and Law, 12, 137-169.
Meissner, C.A., Sporer, S.L., & Susa, K.J. (2008). A theoretical review and metaanalysis of the description-identification relationship in memory for faces. European
Journal of Cognitive Psychology, 20, 414-455.
Meissner, C.M., Tredoux, C.G., Parker, J.F., & MacLin, O. (2005). Investigating the
phenomenological basis for eyewitness decisions in simultaneous and sequential lineups: A dual-process signal detection theory analysis. Memory & Cognition, 33, 783792.
Milne, R. & Bull, R. (1999). Investigative interviewing: Psychology and practice.
Chichester: John Wiley & Sons, Ltd.
Olsson, N., & Juslin, P. (1999). Can self-reported encoding strategy and recognition
skill be diagnostic of performance in eyewitness identification? Journal of Applied
Psychology, 84, 42-49.
36
PACE Code D (2008). Code of practice for the identification of persons by police
officers. Downloaded on 6/12/2009 from http://police.homeoffice.gov.uk/operationalpolicing/powers-pace-codes/pace-code-intro/
Rattner, A. (1988). Convicted but innocent: wrongful conviction and the criminal
justice system. Law and Human Behavior, 12, 283-293.
Read, J.D., Hammersley, R., Cross-Calvert, S., & McFadzen, E. (1989). Rehearsal of
faces and details in action events. Applied Cognitive Psychology, 3, 295-311.
Shapiro, P.N., & Penrod, S.D. (1986). Meta-analysis of facial identification rates.
Psychological Bulletin, 100, 139-156.
Shepherd, J.W. (1983). Identification after long delays. In S.M.A. Lloyd-Bostock, and
B.R. Clifford (eds.), Evaluating witness evidence (pp. 173–187). Chichester: Wiley.
Shepherd, J.W., Davies, G.M., & Ellis, H.D. (1978). How best shall a face be
described? In M. M. Gruneberg, P. E. Morris, & R. N. Sykes (Eds.). Practical aspects
of memory. New York: Academic Press.
Skelton, F. & Frowd, C.D. Facial context and composites. Manuscript in preparation.
Sondhi, V., & Gupta, A. (2007). The quality of suggested memories. International
Journal of Psychology, 42, 116-123.
37
Sörqvist, P., & Eriksson, M. (2007). Effects of training on age estimation. Applied
Cognitive Psychology, 21, 131-135.
Sporer, S.L. (1996). Psychological aspects of person descriptions. In S.L. Sporer, R.S.
Malpass & G. Koehnken (Eds.), Psychological issues in eyewitness identification (pp.
53-86). Hillsdale, NJ: Lawrence Erlbaum.
Steblay, N.M. (1992). A meta-analytic review of the weapon focus effect. Law and
Human Behavior, 16, 413-424.
Tanaka, J.W., & Farah, M.J. (1993). Parts and wholes in face recognition. Quarterly
Journal of Experimental Psychology: Human Experimental Psychology, 46A, 225245.
Valentine, T., & Heaton, P. (1999). An evaluation of the fairness of police line-ups
and video identifications. Applied Cognitive Psychology, 13, S59-S72.
Valentine, T., & Mesout, J. (2009). Eyewitness identification under stress in the
London Dungeon. Applied Cognitive Psychology, 23, 151-161.
Valentine, T., Pickering, A. & Darling, S. (2003). Characteristics of eyewitness
identification that predict the outcome of real line-ups. Applied Cognitive Psychology,
17, 969-993.
38
Wells, G.L. (1985). Verbal descriptions of faces from memory: are they diagnostic of
identification accuracy? Journal of Applied Psychology, 70, 619-626.
Wells, G.L., & Hryciw, B. (1984). Memory for faces: encoding and retrieval
operations. Memory & Cognition, 12, 338-344.
Wells, G.L., Olson, E.A., & Charman, S.D. (2002). The confidence of eyewitnesses in
their identifications from line-ups. Current Directions in Psychological Science, 11,
151-154.
Wells, G. L., & Turtle, J.W. (1988). What is the best way to encode faces? In M.M.
Gruneberg, P. Morris, & R. Sykes (Eds.). Practical aspects of memory: Current
research and issues. Vol. 1. Chichester: Wiley. 163-168.
Yuille, J.C. & Cutshall, J.L. (1986) A case study of eyewitness memory of a crime.
Journal of Applied Psychology, 71, 291-301.
39
LEARNING OUTCOMES
At the end of this chapter you should be able to answer the following questions –
1. Explain how good we are at describing unfamiliar people, objects and events.
2. Describe what the main differences are between familiar and unfamiliar face
recognition, and how this relates to eyewitness memory.
3. Describe how effective facial composites are and what can be done to improve
their effectiveness.
4. Outline why eyewitness evidence can be unreliable.
5. List the main techniques used in the enhanced cognitive interview.
40
CHAPTER 14 – KEY TERMS
Caricatured composite
Cognitive Interview (CI)
Context
Conversational Management (CM)
Cued recall
Description-identification relationship
E-FIT
Encoding
Enhanced Cognitive Interview (ECI)
EvoFIT
External facial features
Eyewitness
Facial composite
Facial configuration
False memory
Familiar face recognition
Feature composite system
Free recall
Group identification
Hair
Holistic cognitive interview
Identification parade
Individual facial features
41
Internal facial features
Lineup
Memory
Miscarriage of justice
Mock witness paradigm
Morphed composite
PEACE interview
Person descriptions
Photofit
Picture matching
PRO-fit
Prototype
Rapport building
Recall
Recognition
Reinstating the context
Repeated recall
Report everything
Retrograde amnesia
Sequential lineup
Simultaneous lineup
Sketch artist
Standard Interview
Structural code
Suspect
42
Unfamiliar face recognition
Victim
Video identification
VIPER (Video Identification Parade Electronic Recording)
Weapon focus
Witness
43
REVIEWS
Bruce, V. (1998). Fleeting images of shade: Identifying people caught on video.
President’s Award lecure. The Psychologist, July, 331-337. Available at
http://www.bps.org.uk
Bruce, V., & Young, A.W. (1986). Understanding face recognition. British Journal of
Psychology, 77, 305-327.
Deffenbacher, K.A., Bornstein, B.H., Penrod. S.D, & McGorty, E.K. (2004). A metaanalytic review of the effects of high stress on eyewitness memory. Law and Human
Behavior, 28, 687-706.
Devlin, P. (1976). Report to the Secretary of State for the Home Department of the
Departmental Committee on Evidence of identification in criminal cases. London:
HMSO.
Ellis, H.D. & Shepherd, J.W. (1996). Face recall methods and problems. In S.L.
Sporer, R.S. Malpass & G. Koehnken (Eds.). Psychological issues in eyewitness
identification. Hillsdale, NJ: Lawrence Erlbaum. Ch. 5.
Frowd, C.D., Bruce, V., & Hancock, P.J.B. (2008). Changing the face of criminal
identification. The Psychologist, 21, 670-672.
44
George, P.A., & Hole, G.J. (2000). The role of spatial and surface cues in the ageprocessing of unfamiliar faces. Visual Cognition, 7, 485-509.
Griffiths, A., & Milne, R. (2005). Will it all end in tiers? Police interviews with
suspect in Britain. In T.Williamson (Ed.), Investigative interviewing. Rights, research,
regulation. Devon:Willan Publishing.
McQuiston-Surrett, D., Malpass, R. S., & Tredoux, C. G. (2006). Sequential vs.
simultaneous line-ups: A review of methods, data, and theory. Psychology, Public
Policy and Law, 12, 137-169.
Meissner, C.A., & Brigham, J.C. (2001). Thirty years of investigating the own-race
bias in memory for faces: A meta-analytic review. Psychology, Public Policy and
Law, 7, 3-35.
Meissner, C.A., Sporer, S.L., & Susa, K.J. (2007). A theoretical review and metaanalysis of the description-identification relationship in memory for faces. European
Journal of Cognitive Psychology, 20, 414-455.
Rattner, A. (1988). Convicted but innocent: wrongful conviction and the criminal
justice system. Law and Human Behavior, 12, 283-293.
Steblay, N.M. (1992). A meta-analytic review of the weapon focus effect. Law and
Human Behavior, 16, 413-424.
45
Wells, G.L., & Olson, E.A. (2003). Eyewitness testimony. Annual Review of
Psychology, 54, 277-295.
Wells, G., Memon, A., & Penrod, S.D. (2007). Eyewitness evidence: improving its
probative value. Psychological sciences in the public interest, 7, 45-75.
46
TEXTS FOR FURTHER READING
A. Baddeley (2004). Human memory: a user’s guide. New York: Firefly Books.
M. Matlin (2009). Cognitive psychology. Wiley.
R. Milne, & R. Bull (1999). Investigative interviewing: Psychology and practice.
Chichester: John Wiley & Sons, Ltd.
S.L. Sporer, R.S. Malpass & G. Koehnken (2006). Psychological issues in eyewitness
identification. Hillsdale, NJ: Lawrence Erlbaum.
T. Williamson (2005). Investigative interviewing. Rights, research, regulation. Devon:
Willan Publishing.
A. Young & H.D. Ellis (1989). Handbook of research on faces. Amsterdam: NorthHolland.
47
CHAPTER 14
SECTION SUMMARIES
14.0
INTRODUCTION

Eyewitnesses help the police bring a criminal to justice.

They give description of crimes, take part in identification procedures,
construct facial composites and give testimony in a court of law.

The police may make a public appeal for information.

When there is good reason to believe that a person has committed an offence,
he or she becomes a suspect in a criminal investigation.
14.1
INFORMATION RECALL

Information gained from an eyewitness is one source of evidence.

Evaluations of real life crimes reveal that witnesses remember information
with a high degree of accuracy and maintain this for long periods of time.

Errors made are about colours of clothing and hair.

Eyewitnesses are generally fairly accurate at judging estimates of height,
weight and age.

The way in which a question is asked may affect recall and may result in a
false memory being created.

Physiological stress when observing an event can result in retrograde amnesia.
48

Attention can be drawn to the presence of a weapon, reducing overall recall.

The cognitive interview is a set of techniques that has been designed to
maximise the quality and quantity of information recall.

An enhanced version of the cognitive interview is even more effective than the
standard version.

The UK police use interviews based on the enhanced cognitive interview and
additional memory-enhancing techniques.

Human faces contain individual facial features and a facial configuration.

Descriptions of faces are valuable to a police investigation and tend to focus
on the exterior facial region.

Facial distinctiveness is an important factor for face recall and recognition.

Information about faces is lost over time, the same as for information in
general.

14.2

Face recall is better under a feature encoding strategy.
PERSON RECOGNITION
There are a range of cues we use to recognise a person but information from
the face is arguably the most important.

We are excellent at recognising faces with which we are familiar.

Familiar faces are recognised more by their internal than their external
features.

Offenders are normally only seen once and therefore eyewitnesses use
unfamiliar face perception to describe and recognition the face.
49

Familiar face recognition is sensitive to a range of effects including lighting,
head pose, expression and context.

The external features exert a greater influence overall for unfamiliar than for
familiar face perception.

The UK police use group identification and lineups procedures with
eyewitnesses.

Lineups can involve real people or videos and are administered sequentially or
simultaneously.

They should be constructed on the basis of an eyewitness’s description of an
offender.

Larger sized lineups are more effective as they protect innocent suspects.

Lineup fairness can be tested using a mock witness paradigm.
14.3

FACE CONSTRUCTION
Eyewitnesses may construct a picture of an offender’s face when the police do
not have a suspect.

There are various methods that have been developed to build the face and
these include sketch artists and ‘feature’ systems.

Modern computerised feature systems contain a large number of individual
facial features for eyewitnesses to select and position on the face.

Research suggests that feature systems are somewhat effective when the target
delay is up to a few hours in duration but not when it is 2 days or more, the
norm for ‘real’ witnesses.
50

There are a range of techniques developed to improve the effectiveness of
facial composites when the delay is long:

Composites from different witnesses may be combined to produce a morphed
composite.

Eyewitnesses may be asked the think about the personality of the offender’s
face as part of an ‘advanced’ cognitive interview.

Identification of a composite (to members of the public) is improved using a
caricaturing procedure.

All these techniques are ineffective if an eyewitness cannot recall the face in
detail.

The EvoFIT system allows a composite to be constructed in situations where
an eyewitness’s recall of the face is very poor.

EvoFIT presents screens of complete faces to allow a composite to be evolved.

The system is much more effective than a feature system when the target delay
is 2 days.
14.4

EVIDENCE AND EYEWITNESSES
There are many types of evidence available to a criminal investigation, but
some of these can be misleading.

There have been many cases of wrongful conviction that have occurred due to
errors of eyewitness identification.

Convictions can be rendered more reliable by combining evidence from
different sources.
51
CHAPTER 14
SELF-TEST QUESTIONS
14.1
INTRODUCTION

Describe the various sources of evidence available in a police investigation.

What is a suspect?
14.1
INFORMATION RECALL

How well do eyewitnesses recall information?

Describe the kinds of inaccuracies that eyewitnesses make.

What factors affect what we recall?

Describe the cognitive interview and say how it has been enhanced.

The UK police use a special kind of interview. What is this called and what
are the techniques it contains?

What is weapon focus?

How does state anxiety affect information recall?

Do we recall different parts of a human face in different ways? If so, how?
14.2

PERSON RECOGNITION
What cues do we use to recognise a person? Which of these is the most
effective?
52

How are familiar and unfamiliar face recognition different to each other and
how does this relate to eyewitnesses?

Describe the different types of identification procedures.

What are the potential problems with constructing lineups and how can these
be overcome?

14.3
What kinds of errors are made when recognising unfamiliar faces?
FACE CONSTRUCTION

What is a facial composite?

Describe the traditional systems that the police use to construct faces with
eyewitnesses.

How effective are composite systems?

When do feature system run into difficulty?

Various developments have improved the effectiveness of traditional feature
composites. What are these?

Describe the alternative approach to construction by individual features.

How effective is one of these ‘holistic’ systems? What has been done to
improve its performance even further?
14.4
EVIDENCE AND EYEWITNESSES

What is the problem with evidence collected at a crime scene?

What happened in the investigation of Laszlo Virag?

Can eyewitness evidence be reliable? If not, why? (also see 14.2.)
53
CHAPTER 14
RESEARCH QUESTIONS
(This section is to be used by teachers/instructors to lead class discussions on
eyewitness research issues.)
14.0

INTRODUCTION
What are the different sources of evidence collected in a criminal
investigation?
14.1
INFORMATION RECALL

What is the difference between recall and recognition?

What type of errors do eyewitnesses make when they recall information?

What do witness’s recall the best about another person’s face and why?

Why might it be possible to accurately describe details of a crime but not
details of a person’s face?

Why is facial distinctiveness an important factor for face perception?

What are the techniques that assist information recall in the PEACE
interview?
14.2
PERSON RECOGNITION
54

What sources of information do we use when we try to recognise a person?
Which one is likely to be the most important?

What is the consequence of becoming familiar with a person’s appearance?

What kinds of errors do eyewitnesses make with face recognition?

There is a consequence of constructing lineups directly from a picture of a
suspect’s face. Describe what this is.

14.3
Suggest good procedures for building lineups.
FACE CONSTRUCTION

What are the three main methods used to construct a facial composite?

What is the ‘cognitive’ approach and why is it effective?

How useful are facial composites?

When building a composite system’s database, why must only a sample of
facial features be taken from each photographic subject?

Why should composite quality be poor when a person constructs a composite
two days after seeing the face?

Which region of the face is important when recognising a composite, and
why?

Why are the recent techniques likely to be effective for improving the
performance of facial composites?
14.4
EVIDENCE AND EYEWITNESSES
55

How reliable is eyewitness evidence?

What methods can be used to improve their effectiveness?
56
CHAPTER 14
FORENSIC ISSUES
(This section is to be used by teachers/instructors to lead class discussions on forensic
issues.)
14.0

INTRODUCTION
Should evidence be collected from an eyewitness in any order? If so, what
might this be?
14.1

INFORMATION RECALL
What normally happens to information in our memory over time, and does this
occur in real crimes?

What is a false memory and how is it created?

Why might it be possible to accurately describe details of a crime but not
details of a person’s face?

Why is hair the easiest to describe?

What techniques are used to maximise the recall of information?
14.2

PERSON RECOGNITION
What type of face perception do eyewitnesses normally engage in and why
might this be a problem?
57

What is a major source of error in person recognition?

What are good procedures for constructing lineups?

Why is it best to use lineups with a large number of foils?

What problem occurs if an offender substantially changes their hair between
the crime and a lineup?
14.3

FACE CONSTRUCTION
Which techniques are used in the cognitive interview with eyewitnesses for
obtaining a description of an offender’s face? Which are not used?

What do the Frowd et al. (2005a) naming data imply for police practice?

When is it permissible for a composite to be constructed and when is it not?

What are the implications of an offender’s face being distinctive?

In what way can good quality composites be constructed from traditional
systems, and what techniques are available to improve their performance?

14.4

Does an eyewitness need to have good face recall to use EvoFIT?
EVIDENCE AND EYEWITNESSES
What is the main reason for miscarriages of justice?
58
CHAPTER 14
ACTIVITY BOX 14.1
Have a go at describing the appearance of a good friend of yours from memory. First
try to visualise his or her face in your mind and then describe as much as you can. If
you wish, write down what you remember using the following verbal description
sheet. When done, compare your pattern of recall with section 14.1.4 that describes
general findings from the face recall literature.
Verbal Description Sheet
Overall observations
Face shape
Hair
Brows
Eyes
59
Nose
Mouth
Ears
60
FOCUS POINT 14.2.1
In 1969, Laszlo Virag was convicted of theft of money from parking meters in
Liverpool and Bristol, and with the subsequent use of a firearm while attempting to
escape arrest. Laszlo’s conviction was based solely on the evidence of eight
eyewitnesses who identified him from a lineup. He was sentenced to 10 years in
prison. Two years later, as part of a separate investigation, fingerprint and other
evidence were found to implicate another person, Georges Payen, and Laszlo was
exonerated and compensated for his time served in prison. [FIGURE Focus Point
14.2.1]
The case of Laszlo Virag is just one of many unfortunate convictions that have
been overturned and innocent people set free. The case against him was based on
eyewitness testimony that appeared to be both accurate and convincing. The
observers felt confident in their judgement, and so clearly feelings cannot always be
relied upon, even though they may sound convincing to a jury (Wells, Olson &
Charman, 2002). One of the eyewitnesses was even reported to have spent several
minutes in a hotel bar with the accused but still mistakenly picked out Virag in a
lineup. Virag and Payen share a passing but not a striking resemblance to each other.
A formal inquiry into such miscarriages of justice was carried out by Lord Devlin in
the 1970s with the conclusion that eyewitness evidence was unreliable. There are
many other cases involving honest but nonetheless incorrect identification; for further
examples, see Davies & Griffiths (2008) or Rattner (1988).
61
TABLE 14.1.2
The following table lists the main stages of the enhanced cognitive interview (ECI)
developed by Fisher et al. (1987).
1. Rapport building
2. Describe the aims of the interview
3. Context reinstatement and free recall
4. Open questions (cued recall)
5. Repeated recall (different orders and perspectives)
6. Summary
7. Close
FIGURE 14.0.
A Wanted Poster released by Northants police as part of Operation Mallard. The
criminal investigation involved a series of sexual assaults in Southern England in the
late 1990s and January 2000.
62
63
FIGURE 14.1.1
The influence of people’s perception of vehicle speed following the verb used in
Loftus & Palmer (1974). Average estimates of speed increased from 31.8 mph using
‘contacted’ to 40.5 mph using ‘smashed’.
Mean speed estimate (mph)
50.0
40.0
30.0
20.0
10.0
0.0
Contacted
Hit
Bumped
Collided
Smashed
Verb
Redrawn from Table 1, Loftus & Palmer (1974).
64
FIGURE 14.1.2
The figure illustrates the influence of interviewing on information recall. Correct
information recalled was found to be best for the cognitive interview and worst for the
Details recalled (percent correct)
standard interview.
50
40
30
20
10
0
Standard
Cognitive
Hypnosis
Type of interview
Redrawn from Table 1, Geiselman et al. (1985).
65
FIGURE 14.2
Example stimuli from Burton et al. (1999). Participants were presented with (left to
right) un-edited CCTV footage, or footage with the body obscured or the head
obscured. Also included was a condition where gait was disrupted. The greatest
disturbance to face recognition occurred with the face obscured (far right image),
indicating the importance of this part of a person’s appearance.
Reprint of Fig. 3 from Burton et al. (1999) with permission. [Editor: please obtain
permission]
66
FIGURE 14.2.1
The graph demonstrates that the changes made to a face between study and test affect
our ability to recognise the face. Participants in Bruce (1982), Experiment 1, saw the
same picture at study and test (far left bar), a change in either pose or expression
(centre bar), or a change in both pose and expression (far right bar).
Accuracy (percent correct)
100
80
60
40
20
0
None (same picture)
1 Change
2 Changes
Change in face between study and test
67
FIGURE 14.3.2.A
Example composites from Frowd et al. (2005b). All were constructed from different
peoples’ memories. The identity is the same throughout: can you guess who it is?
The answer is listed in section 14.5. From left to right, they were constructed from EFIT, PRO-fit, Photofit, Sketch and EvoFIT.
[Editor: if a picture of Michael Owen can be obtained, this could be included here for
reference. I do not have one with copyright clearance.]
68
FIGURE 14.3.2.B
The graph demonstrates how effective composites are when the delay-to-construction
is fairly short (three to four hours). Frowd et al. (2005b) found composite naming to
be only fairly good (less than about 20% correct). The two main UK software
systems, E-FIT and PRO-fit, were equivalent and better than the older Photofit.
(Note that 100% on this scale would mean that all composites from a particular
system would have been correctly named by all participants.)
69
FIGURE 14.3.3
Example stimuli used in Frowd et al. (2007a): internal and external composite
features, and the corresponding complete composite. In the study, participants
matched complete composites, or one of these part-composite images, to target
photographs. The internal feature composites were not matched as well as the other
types of image, indicating poor construction of the central facial region.
70
FIGURE 14.3.4.A
Example composites (top row) constructed by participants in Bruce et al. (2002). The
morphed image of these individual composites is shown on the bottom left; the
researchers demonstrated that this is an effective probe for the target face (in this case,
the image on the bottom right).
Reprinted from Figure 4, Bruce et al. (2002) with permission. [Editor: please obtain
permission]
71
FIGURE 14.3.4.B
Caricaturing a composite of the former Prime Minister, Tony Blair. Shown are
negative and positive caricature levels: -50%, -25%, 0%, 25% and 50%
exaggerations. The image in the centre (0%) is the original composite constructed by
a participant-witness. A sequence of such images (with 21 frames) is valuable for
face recognition. An animated example may be seen at www.uclan.ac.uk/caricature.
72
FIGURE 14.3.4.C
An example array of EvoFIT faces as would be presented to eyewitnesses. The
external parts of the face have been blurred, to allow them to focus on the important,
central region. At the end of evolving, the blur is disabled to allow the entire image to
be seen clearly.
73
FIGURE 14.3.4.D
An example of the ageing (top row) and pleasantness (bottom row) scales in EvoFIT.
At the end of evolving, eyewitnesses are given the opportunity to manipulate their
final face using these and other ‘holistic’ scales. The centre image is veridical,
negative changes on the scale are to the left and positive changes are to the right. Can
you guess the identity of the evolved face being manipulated here? The answer is
listed in section 14.5.
74
Figure for FOCUS POINT 14.2.1
The case of Laszlo Virag. The photograph on the left is of Laszlo Virag, convicted in
1969 of a series of bank robberies in NW England. The image on the right is of
Georges Payen who was later found to be responsible for the offences. The images
share a passing but not a striking resemblance to each other.
Reproduced from Devlin (1986) with permission. [Editor: please obtain permission]
75
14.5
ANSWERS
Composites in Figure 14.3.2.A are of the UK footballer, Michael Owen. The
images in Figure 14.3.4.D are based on a composite of Simon Cowell.
76
Acknowledgement
The author would like to thank Dr Faye Skelton, University of Central Lancashire,
Preston, for her insightful comments on a draft of Chapter 14, and Emma Walker for
proof reading a near-final version.
77